Machine Learning and TSN with NXP's i.MX 8M Plus
Вставка
- Опубліковано 7 жов 2020
- In this demo we showcase the Machine Learning capabilities of NXP's i.MX 8M Plus as well as the improved performance in motion control thanks to TSN communication.
Follow us on:
LinkedIn - / nxp-semiconductors
Twitter - / nxp
facebook - / nxpsemi - Наука та технологія
The fast Walsh Hadamard transform is a collection of orthogonal dot products. If you apply parametric activation functions after you have a very fast neural layer. Since the dot products are summary measures the system is highly statistical in nature. Practically you might do a random projection before the first layer and apply a WHT at the end as a pseudo-readout layer. Maybe f(x)=a.x x>=0, f(x)=b.x x